Sparse Representation Classification Beyond L1 Minimization and the Subspace Assumption